Gemini 2.5 Pro vs Gemini 2.5 Flash
Lado a lado. Cada métrica. Cada benchmark.
| Tipo | Gemini 2.5 Pro | Gemini 2.5 Flash |
|---|---|---|
| Provider |  Google DeepMind | Google DeepMind |
| pontuação média | 56.2 | 40.0 |
| Preço de entrada | $1.25 | $0.30 |
| Preço de saída | $10.00 | $2.50 |
| Janela de contexto | 1.0M tokens (~524 books) | 1.0M tokens (~524 books) |
| Lançado em | 2025-06-17 | 2025-06-17 |
| Código aberto | Proprietary | Proprietary |
Pontuações de benchmark
25 benchmarks · Gemini 2.5 Pro: 22, Gemini 2.5 Flash: 2
| Benchmark | Categoria | Gemini 2.5 Pro | Gemini 2.5 Flash |
|---|---|---|---|
| Aider polyglot | coding | 83.1 | 47.1 |
| ARC-AGI | reasoning | 41.0 | 32.3 |
| ARC-AGI-2 | reasoning | 4.9 | 2.5 |
| AudioMultiChallenge | knowledge | 46.9 | 40.0 |
| AudioMultiChallenge — Text Output | knowledge | 46.9 | 40.0 |
| Balrog | knowledge | 43.3 | 33.5 |
| Chatbot Arena Elo — Overall | arena | 1448.2 | 1411.0 |
| DeepResearch Bench | knowledge | 49.7 | 29.2 |
| Fiction.LiveBench | knowledge | 91.7 | 47.2 |
| FrontierMath-2025-02-28-Private | math | 14.1 | 4.8 |
| FrontierMath-Tier-4-2025-07-01-Private | math | 4.2 | 4.2 |
| GeoBench | knowledge | 81.0 | 73.0 |
| HELM — GPQA | knowledge | 74.9 | 39.0 |
| HELM — IFEval | language | 84.0 | 89.8 |
| HELM — MMLU-Pro | knowledge | 86.3 | 63.9 |
| HELM — Omni-MATH | math | 41.6 | 38.4 |
| HELM — WildBench | reasoning | 85.7 | 81.7 |
| HLE | knowledge | 17.7 | 7.7 |
| Lech Mazur Writing | knowledge | 86.0 | 76.5 |
| OTIS Mock AIME 2024-2025 | math | 84.7 | 73.0 |
| SimpleBench | reasoning | 54.9 | 29.4 |
| Terminal Bench | coding | 32.6 | 17.1 |
| The Agent Company | agentic | 30.3 | 41.1 |
| VPCT | knowledge | 19.6 | 7.0 |
| WeirdML | coding | 54.0 | 41.0 |